Write your first eval

A case, a check and a score, in under twenty lines.

LAST UPDATED

Oct 3, 2026

An eval is a function that returns a score between zero and one.

A case

export default tracewell.eval('refund-edge-12', {
  input: { ticket: 'I was charged twice...' },
  check: (out) => out.includes('refund') ? 1 : 0,
})

export default tracewell.eval('refund-edge-12', {
  input: { ticket: 'I was charged twice...' },
  check: (out) => out.includes('refund') ? 1 : 0,
})

export default tracewell.eval('refund-edge-12', {
  input: { ticket: 'I was charged twice...' },
  check: (out) => out.includes('refund') ? 1 : 0,
})

npx tracewell eval run

npx tracewell eval run

npx tracewell eval run

The result links to the run so you can open the trace of any failing case.

← PREVIOUS